cvlab-ai

KO | EN

gitlite — search

#javascript #react #llm #machine-learning #rust #deep-learning #docker #macos #windows #kubernetes #awesome-list #vue #security #ios

(★ 10)

Computer Vision Laboratory

cvlab-ai Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# 🎥 Computer Vision Lab (CSET340) Assignments

An engaging course repository for computer vision assignments, focusing on practical implementation of image processing, feature detection, object recognition, and deep learning techniques using Python and OpenCV.

![Python](https://img.shields.io/badge/Python-3.9+-blue.svg)
![OpenCV](https://img.shields.io/badge/OpenCV-4.8+-green.svg)
![PyTorch](https://img.shields.io/badge/PyTorch-2.0+-red.svg)
![License](https://img.shields.io/badge/License-MIT-yellow.svg)
![Maintenance](https://img.shields.io/badge/Maintenance-Active-brightgreen.svg)

## 📖 Table of Contents
- [Core Components](#-core-components)
- [Technical Requirements](#-technical-requirements)
- [Getting Started](#-getting-started)
- [Assignments](#-assignments)
    * [Experiment 1](#-Experiment-1)
    * [Experiment 8](#-Experiment-8)
    * [Experiment 9](#-Experiment-9)
- [Resources](#-resources)

## 🌟 Core Components

### 📸 Image Processing Fundamentals
- **Basic Operations**
  - Image manipulation
  - Color space conversions
  - Filtering techniques
  - Histogram analysis
- **Advanced Techniques**
  - Edge detection
  - Morphological operations
  - Image enhancement
  - Frequency domain processing

### 🔍 Feature Detection & Recognition
- **Feature Extraction**
  - SIFT/SURF implementations
  - Corner detection
  - Blob detection
  - Template matching
- **Pattern Recognition**
  - Feature matching
  - Object detection
  - Face recognition
  - Scene classification

### 🧠 Deep Learning Approaches
- **Neural Networks**
  - CNN architectures
  - Transfer learning
  - Model training
  - Performance optimization
- **Modern Architectures**
  - ResNet
  - YOLO
  - U-Net
  - Transformers

## 🔧 Technical Requirements

### System Setup
- **Python Environment**
  - Python 3.9+
  - pip or conda
  - Virtual environment
  - Git
- **Required Libraries**
  - OpenCV 4.8+
  - NumPy 1.21+
  - PyTorch 2.0+
  - Matplotlib 3.5+

### Dependencies
```txt
# requirements.txt
opencv-python>=4.8.0
numpy>=1.21.0
torch>=2.0.0
matplotlib>=3.5.0
scikit-image>=0.19.0
pillow>=9.0.0
jupyter>=1.0.0
```

## 🚀 Getting Started

### Setup Instructions
```bash
# Clone the repository
# git clone https://github.com/university/cv-course.git
# cd cv-course

# Create virtual environment
# python -m venv venv
# source venv/bin/activate  # Windows: venv\Scripts\activate

# Install dependencies
# pip install -r requirements.txt

# Verify installation
# python -c "import cv2; print(cv2.__version__)"
```

## ⭐ Grading

### Evaluation Criteria
| Component | Weight | Description |
|-----------|---------|------------|
| Implementation | Code correctness & efficiency |

| Results | Output quality & analysis |

## 🤝 Contributing

### Guidelines
1. Follow PEP 8 style guide
2. Document all functions
3. Include unit tests
4. Maintain clean commit history

## 🚀 Assignments

## 🌟 Experiment 1
Task-1: Perform following operations on image- 
-Image Resizing: Resizing involves changing the dimensions of an image, either by scaling it up or down.
-Image resizing (interpolation methods)
- Linear
- Nearest Neighbors
- Polynomial
- Image Blurring: Blurring is used to reduce image detail, suppress noise, or create artistic effects. Common techniques include:
1. Image blurring
2. Box blurring
3. Gaussian blurring
4. Adaptive blurring

Task-2: Apply Machine Learning Algorithm and find the model accuracy based on K fold Cross Validation with (80-20 train-test split).  
1. Use MNIST dataset
2. Use any two of the following algorithms-
3. Naive Bayesian or its variant.
4. Support Vector Machine (SVM) or its variant
5. Decision Trees/ Random Forest.
6. AdaBoost or other ensemble algorithms.
7. Artificial Neural Networks (NN) or its variant.
- Results should be obtained on following parameters-
1. Accuracy
2. Precision (Positive Predictive Value)
3. Recall (Sensitivity)
4. F-Measure
5. Confusion Matrix
6. ROC
7. AUC
- Appendix:
About MNIST :-
- The MNIST dataset stands for "Modified National Institute of Standards and Technology"
- The dataset contains a large collection of handwritten digits that is commonly used for training various image processing systems.
- The dataset was created by re-mixing samples from NIST's original datasets, which were taken from American Census Bureau employees and high school students.
- It contains 60,000 training images and 10,000 testing images, each of which is a grayscale image of size 28x28 pixels.
- Number of Instances: 70,000 images
- Number of Attributes: 784 (28x28 pixels)
- Target: Column represents the digit (0-9) corresponding to the handwritten image
- Pixel 1-784: Each pixel value (0-255) represents the grayscale intensity of the corresponding pixel in the image.
- The dataset is divided into two main subsets:
- Training Set: Consists of 60,000 images along with their labels, commonly used for training machine learning models.
- Test Set: Contains 10,000 images with their corresponding labels, used for evaluating the performance of trained models.
- Link:- https://www.kaggle.com/datasets/hojjatk/mnist-dataset
- Note:- Use sklearn, pyspark, or any other ML library for applying the ML algorithms.
- Load the dataset in sklearn using ‘load_digits’.
- Load the dataset in pyspark using 'spark.read.csv()”

## 🌟 Experiment 8
### Task 1.1- Blob detection and Image Quality Enhancement
- Blob detection:- Blob detection is a basic method in computer vision used to locate areas of interest in a picture.
- Apply the three different blob detection techniques (LoG, DoG, HoG) separately on the same image
### Task 1.2- Image quality enhancement (Perform this using the following techniques)
- Adjusting brightness and contrast
- Sharpening images
- Removing noise from images
- Enhancing color in images
- Image resizing and scaling
- Inverse Transform
- Equalizing histograms
- Super-resolution
- Color correction
### Task 2.1- Image Classification using Resnet network on Cifar 100
- Compare the performance of Resnet 18 and Resnet 34  an image classification task.
- Use CIFAR-100, a common dataset for image classification.
- Analyse model accuracy, loss, and inference time on a dataset.
### Task 2.2- Meta learning approaches for image classification on MNIST dataset
- Zero Shot
- One Shot
- Few Shot
---

## 🌟 Experiment 9
### Task 1- Interest Point Detection, Feature Matching and Contour Detection
* Interest Point Detection:-  Apply SIFT (Scale Invariant Feature Transform) Detector function using cv.SIFT_create()
* Feature Matching:- Feature matching is a fundamental technique in computer vision and image processing that involves finding correspondences between features detected in different images.
* Use methods namely ORB (Oriented FAST and Rotated BRIEF) and BFMatcher (Brute-Force Matcher).
* Contour Detection with Custom Seeds:- Contour represents the outline or boundary of an object, connecting continuous points of similar intensity or color.
* In image processing, edges represent abrupt changes in brightness or color, while contours are closed curves that outline the shape or form of an object, often derived from edges.
* Functions like markers, watershed, with some additional color placement functions can be used.

### Task 2- Image Restoration using Autoencoder Model
#### Step 1: Load & Preprocess Data
*	Load dataset (e.g., MNIST, CIFAR-10).
*	Normalize pixel values between [0,1].
*	Resize images if necessary.
* 	Convert to grayscale (if required).
#### Step 2: Simulate Degradation
*	Introduce noise (Gaussian noise, salt & pepper noise).
*	Apply blur (motion blur, Gaussian blur).
*	Add compression artifacts if needed.
#### Step 3: Build the Autoencoder Model
*	Encoder: Uses convolutional layers to extract features and compress image representation.
*	Decoder: Reconstructs the image from the compressed representation using up-sampling and convolutional layers.
#### Step 4: Train the Model
*	Train autoencoder using degraded images as input and original images as target output.
*	Use Mean Squared Error (MSE) as the loss function.
*	Optimize with Adam optimizer.
*	Validate on test data.
#### Step 5: Image Restoration & Evaluation
*	Predict restored images using the trained model.
*	Compare original, degraded, and restored images.
*	Evaluate performance using PSNR (Peak Signal-to-Noise Ratio)

### Generative Adversarial Networks on MNIST and Cifar-10 datasets
- Take any GAN variant and generate 10 different digits of MNIST dataset and 10 different classes of cifar-10 dataset and plot the discriminator vs adversarial loss for both.
- Min epoch – 50

### Mobilenet based Image Classification
- Apply Mobilenet on dogs-breed dataset as part of image classification task.
- Compare the three Versions V1,V2 and V3 performance under same hyperparameter setting

### Unet for Image Segmentation
- Task is to first train the  U-Net model on the Oxford-IIIT Pet Dataset or the brain tumor T1-Lighted CE-MRI image dataset to perform foreground-background segmentation and evaluate the segmentation performance using IoU (Intersection over Union) and Dice Coefficient.

*Last Updated: February 2025*

cvlab-ai

cvlab-ai Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

cvlab-ai

cvlab-ai Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README